regen-network · CShear · Feb 18, 2026 · Feb 18, 2026
diff --git a/README.md b/README.md
@@ -79,6 +79,7 @@ This repository contains the design specifications and implementation plans for
 
 <!-- AUTOGENERATED: run node scripts/build-mechanism-index.mjs -->
 - [m010 — Reputation Signal (v0 advisory)](mechanisms/m010-reputation-signal/)
+- [m011 — Marketplace Curation & Quality Signals](mechanisms/m011-marketplace-curation/)
 
 <!-- END MECHANISMS INDEX -->
 

diff --git a/docs/MECHANISM_CONSUMERS.md b/docs/MECHANISM_CONSUMERS.md
@@ -18,3 +18,23 @@ This document maps **mechanism IDs** to known **consumers** (agents, digests, sc
 - Heartbeat replay runner: `scripts/replay-m010.mjs` (regen-heartbeat)
 - Heartbeat stub runner: `scripts/stub-run-signal-agent.mjs` (regen-heartbeat)
 - Heartbeat validator: `scripts/validate-signal-agent.mjs` (regen-heartbeat)
+
+## m011 — Marketplace Curation & Quality Signals
+**Canonical spec**
+- `mechanisms/m011-marketplace-curation/SPEC.md`
+
+**Outputs**
+- KPI JSON block schema: `mechanisms/m011-marketplace-curation/schemas/m011_kpi.schema.json`
+- Quality score schema: `mechanisms/m011-marketplace-curation/schemas/m011_quality_score.schema.json`
+- Collection lifecycle schema: `mechanisms/m011-marketplace-curation/schemas/m011_collection.schema.json`
+
+**Datasets (deterministic)**
+- Replay fixtures: `mechanisms/m011-marketplace-curation/datasets/fixtures/v0_sample.json`
+- Collection challenges: `mechanisms/m011-marketplace-curation/datasets/fixtures/v0_collection_sample.json`
+
+**Known consumers**
+- AGENT-003: Autonomous quality scoring, price monitoring, collection monitoring
+- KOI MCP: methodology metadata analysis via `resolve_entity`
+- Ledger MCP: batch metadata and trade queries
+- x/ecocredit: batch, class, project data; marketplace sell orders
+- Heartbeat: KPI metrics in weekly digest
diff --git a/mechanisms/m011-marketplace-curation/README.md b/mechanisms/m011-marketplace-curation/README.md
@@ -0,0 +1,35 @@
+# m011 — Marketplace Curation & Quality Signals
+
+Quality signals, curated collections, and price discovery for the credit marketplace. Curators stake REGEN to create vetted credit batch collections and earn trade fee shares. AGENT-003 provides automated 7-factor quality scoring.
+
+## What it outputs
+
+- **Quality score** (0–1000): 7-factor weighted assessment (project reputation 0.25, class reputation 0.20, vintage freshness 0.15, verification recency 0.15, seller reputation 0.10, price fairness 0.10, additionality confidence 0.05)
+- **Confidence** (0–1000): Data availability across 7 input signals
+- **KPI block**: Collection counts by status, quality score distribution, trade volume through collections, curation economics, challenge rate
+
+## What it does not do in v0
+
+- No on-chain curation bonds or collection contracts
+- No automated trade fee splitting to curators
+- No on-chain challenge mechanism
+- Agent quality scores are advisory only — published in digest
+
+## Scoring formula
+
+```
+score = 0.25 × project_reputation + 0.20 × class_reputation + 0.15 × vintage_freshness
+      + 0.15 × verification_recency + 0.10 × seller_reputation + 0.10 × price_fairness
+      + 0.05 × additionality_confidence
+```
+
+## How to reference
+
+- **Canonical spec**: `mechanisms/m011-marketplace-curation/SPEC.md`
+- **Schemas**: `mechanisms/m011-marketplace-curation/schemas/`
+- **Reference implementation**: `mechanisms/m011-marketplace-curation/reference-impl/`
+
+## Replay datasets
+
+- `datasets/fixtures/v0_sample.json` — Credit batches across multiple classes with quality scores
+- `datasets/fixtures/v0_collection_sample.json` — Collection lifecycle scenarios including challenges
diff --git a/mechanisms/m011-marketplace-curation/SPEC.md b/mechanisms/m011-marketplace-curation/SPEC.md
@@ -0,0 +1,284 @@
+# M011 — Marketplace Curation & Quality Signals
+
+## 0. Header
+
+| Field | Value |
+|-------|-------|
+| **Mechanism ID** | m011 |
+| **Name** | Marketplace Curation & Quality Signals |
+| **Status** | Draft — v0 advisory |
+| **Owner** | Regen Network core / community |
+| **Scope** | Credit marketplace quality scoring, curated collections, and price discovery |
+
+---
+
+## 1. Problem
+
+The existing credit marketplace lacks quality differentiation. Buyers cannot easily distinguish high-integrity credits from low-quality ones. Without quality signals, curated collections, or price discovery mechanisms, the marketplace suffers from information asymmetry that depresses prices for high-quality credits and erodes buyer confidence. Curators who invest effort in vetting credits have no economic incentive mechanism.
+
+---
+
+## 2. Target actor and action
+
+| Actor | Action |
+|-------|--------|
+| **Curator** | Creates collections of vetted credit batches, bonds REGEN, earns trade fee share |
+| **AGENT-003** | Computes quality scores for credit batches; monitors pricing; flags quality violations |
+| **Buyer** | Purchases credits through curated collections (higher confidence) |
+| **Seller** | Lists credits on marketplace (listing fee) |
+| **Challenger** | Disputes curator inclusion of low-quality batches |
+
+---
+
+## 3. Signal definition
+
+### Quality Score
+
+The agent produces a quality assessment for each credit batch:
+
+| Field | Type | Range | Description |
+|-------|------|-------|-------------|
+| `score` | integer | 0–1000 | 7-factor weighted quality composite |
+| `confidence` | integer | 0–1000 | Data availability indicator |
+| `factors` | object | — | Individual factor scores |
+
+---
+
+## 4. Evidence inputs
+
+| Input | Source | Required |
+|-------|--------|----------|
+| Project reputation score | M010 reputation signal | Preferred |
+| Credit class reputation score | M010 reputation signal | Preferred |
+| Seller reputation score | M010 reputation signal | Optional |
+| Batch start date (vintage) | x/ecocredit batch metadata | Yes |
+| Last verification timestamp | Most recent MsgCreateBatch or attestation | Preferred |
+| Listing price | x/ecocredit/marketplace sell order | If listed |
+| Class price median | AGENT-003 price tracking | Optional |
+| Methodology metadata IRI | Credit class metadata | Optional |
+
+---
+
+## 5. Scoring function
+
+### Quality score
+
+```
+score = 0.25 × project_reputation
+      + 0.20 × class_reputation
+      + 0.15 × vintage_freshness
+      + 0.15 × verification_recency
+      + 0.10 × seller_reputation
+      + 0.10 × price_fairness
+      + 0.05 × additionality_confidence
+```
+
+| Factor | Weight | Description | Range |
+|--------|--------|-------------|-------|
+| `project_reputation` | 0.25 | M010 reputation score for the project | 0–1000 |
+| `class_reputation` | 0.20 | M010 reputation score for the credit class | 0–1000 |
+| `vintage_freshness` | 0.15 | Linear decay: 1000 at issuance, 0 at 10 years | 0–1000 |
+| `verification_recency` | 0.15 | Linear decay: 1000 at last verification, 0 at 3 years | 0–1000 |
+| `seller_reputation` | 0.10 | M010 reputation score for the seller address | 0–1000 |
+| `price_fairness` | 0.10 | 1000 at class median, decaying by deviation (0 at ±50%) | 0–1000 |
+| `additionality_confidence` | 0.05 | Methodology additionality assessment from credit class metadata | 0–1000 |
+
+All factors are clamped to [0, 1000]. Final score is rounded to the nearest integer.
+
+### Confidence
+
+Confidence reflects data availability across seven boolean signals:
+
+| Signal | Check |
+|--------|-------|
+| `project_reputation_available` | Project has M010 reputation score |
+| `class_reputation_available` | Credit class has M010 reputation score |
+| `seller_reputation_available` | Seller has M010 reputation score |
+| `vintage_known` | Batch start date is available |
+| `verification_date_known` | Last verification timestamp available |
+| `price_data_available` | Listing price and class median available |
+| `methodology_available` | Credit class methodology metadata accessible |
+
+`confidence = round(count(true signals) / 7 × 1000)`
+
+### Price fairness calculation
+
+```
+median = class_batch_price_median(credit_type, vintage_year)
+deviation = |listing_price - median| / median
+price_fairness = max(0, round((1.0 - deviation * 2) × 1000))
+```
+
+At median → 1000; 25% deviation → 500; ≥50% deviation → 0.
+
+---
+
+## 6. State machine
+
+### Collection lifecycle
+
+```
+PROPOSED → ACTIVE → UNDER_REVIEW → ACTIVE (curator wins)
+                                  → SUSPENDED (challenger wins) → ACTIVE (top-up)
+                                                                → CLOSED
+         → CLOSED (curator exit, no pending challenges)
+```
+
+| Transition | Trigger | Guard |
+|------------|---------|-------|
+| → PROPOSED | `curator.create_collection(name, criteria, bond)` | bond ≥ min_curation_bond |
+| PROPOSED → ACTIVE | `activation_delay(48h) AND no_challenge` | — |
+| ACTIVE → ACTIVE (add) | `curator.add_to_collection(batch_denom)` | meets criteria, quality ≥ min_quality_score |
+| ACTIVE → ACTIVE (remove) | `curator.remove_from_collection(batch_denom)` | batch in collection |
+| ACTIVE → UNDER_REVIEW | `challenger.challenge_inclusion(evidence) OR agent.flag_violation` | deposit ≥ challenge_deposit |
+| UNDER_REVIEW → ACTIVE | `challenge.resolved(CURATOR_WINS)` | challenger deposit slashed |
+| UNDER_REVIEW → SUSPENDED | `challenge.resolved(CHALLENGER_WINS)` | batch removed, bond slashed |
+| SUSPENDED → ACTIVE | `curator.top_up_bond() AND suspension_expired` | bond ≥ min_curation_bond |
+| SUSPENDED → CLOSED | `timeout(top_up_window) OR curator.close()` | remaining bond refunded |
+| ACTIVE → CLOSED | `curator.close_collection()` | no pending challenges, unbonding period |
+
+---
+
+## 7. Token flows
+
+### Collection creation
+
+```
+Curator ──(curation_bond: 1000 REGEN)──→ Bond Pool
+```
+
+### Trade through collection
+
+```
+Buyer ──(purchase_price)──→ Marketplace
+Marketplace ──(price − trade_fee − curation_fee)──→ Seller
+Marketplace ──(curation_fee: 0.5% of trade)──→ Curator
+Marketplace ──(trade_fee)──→ Community Pool (via M013)
+```
+
+### Challenge resolution — Challenger wins
+
+```
+Bond Pool ──(bond × slash_percentage)──→ slash
+  slash ──(50%)──→ Challenger (reward)
+  slash ──(50%)──→ Community Pool
+```
+
+### Challenge resolution — Curator wins
+
+```
+Challenge Deposit ──(100%)──→ Community Pool (challenger loses deposit)
+```
+
+### Curator exit
+
+```
+Bond Pool ──(remaining bond)──→ Curator (after unbonding_period, 14 days)
+```
+
+### Governance parameters
+
+| Parameter | Default | Range |
+|-----------|---------|-------|
+| `min_curation_bond` | 1,000 REGEN | 100–10,000 |
+| `listing_fee` | 10 REGEN | 0–100 |
+| `curation_fee_rate` | 0.5% (50 bps) | 0–2% |
+| `challenge_deposit_amount` | 100 REGEN | 10–1,000 |
+| `slash_percentage` | 20% of bond | 5–50% |
+| `challenge_reward_share` | 50% of slash | 25–75% |
+| `activation_delay` | 48 hours | 24h–7d |
+| `unbonding_period` | 14 days | 7–30 days |
+| `bond_top_up_window` | 7 days | 3–14 days |
+| `min_quality_score` | 300 | 100–500 |
+| `max_collections_per_curator` | 5 | 1–20 |
+| `quality_refresh_interval` | 24 hours | 6h–7d |
+
+---
+
+## 8. Featured batches
+
+AGENT-003 can recommend batches for featured placement:
+
+| Criteria | Threshold |
+|----------|-----------|
+| Quality score | ≥ 800 (top 20%) |
+| Pending challenges | None |
+| Active sell orders | > 0 |
+
+Featured status lasts 7 days (renewable). Agent recommends (Layer 1), curator confirms (Layer 2).
+
+---
+
+## 9. Security invariants
+
+1. **Bond Conservation**: `bond_pool.balance = sum(active_collection_bonds) + sum(pending_refunds)` at all times.
+2. **Criteria Enforcement**: Every batch in a collection must satisfy the collection's `CurationCriteria` at addition time; AGENT-003 re-verifies daily for drift.
+3. **Curator Authority**: Only the collection curator can add/remove batches; agents cannot modify collection membership directly.
+4. **Slash Cap**: Cumulative slashing cannot exceed the curator's total bond. If `bond_remaining < min_curation_bond` after a slash, the collection transitions to SUSPENDED. If the bond is not topped up within `bond_top_up_window`, the collection then transitions to CLOSED.
+5. **No Self-Challenge**: Curator cannot challenge own collection; challenger cannot be seller of challenged batch.
+6. **Fee Integrity**: Curation fees only on trades through a collection; direct trades incur no curation fee.
+7. **Quality Score Immutability**: Scores are append-only (superseded, never edited); full history preserved.
+8. **Collection Uniqueness**: Each `(collection_id, batch_denom)` pair is unique; a batch can belong to multiple collections.
+
+---
+
+## 10. Attack model
+
+| Attack | Mitigation |
+|--------|-----------|
+| **Low-quality curation** | Challenge mechanism slashes curator bond; AGENT-003 flags quality violations |
+| **Sybil curators** | min_curation_bond (1000 REGEN) makes Sybil attacks expensive |
+| **Challenge spam** | Challenge deposit (100 REGEN) lost if challenge fails |
+| **Price manipulation** | AGENT-003 detects outlier pricing (z-score > 2.5) |
+| **Wash trading for fees** | Fee rate is small (0.5%); on-chain detection via AGENT-003 |
+| **Quality score gaming** | 7-factor scoring with diverse inputs; immutable score history |
+| **Curator front-running** | Activation delay (48h) prevents immediate inclusion of batches |
+| **Bond grinding** | max_collections_per_curator limits; unbonding_period prevents rapid cycling |
+
+---
+
+## 11. Integration points
+
+| System | Integration |
+|--------|-------------|
+| **M010 Reputation** | Project, class, and seller reputation scores as quality inputs |
+| **M013 Fee Routing** | Trade fees follow M013 distribution model |
+| **AGENT-003** | Autonomous quality scoring, price monitoring, collection monitoring |
+| **x/ecocredit** | Batch metadata (vintage, class, project), marketplace sell orders |
+| **x/ecocredit/basket** | Basket token quality scoring (constituent analysis) |
+| **KOI MCP** | Methodology metadata analysis, additionality confidence |
+| **Ledger MCP** | Bond balance queries, trade volume tracking |
+| **Heartbeat** | KPI metrics published in weekly digest |
+
+---
+
+## 12. Acceptance tests
+
+1. **Collection lifecycle**: Curator bonds → creates collection → adds batches → earns trade fees → closes → bond refunded.
+2. **Quality scoring**: AGENT-003 scores batch → score meets criteria → batch eligible for collection inclusion.
+3. **Challenge — challenger wins**: Challenger deposits → challenge resolved in challenger's favor → batch removed → curator bond slashed → challenger rewarded.
+4. **Challenge — curator wins**: Challenge resolved in curator's favor → challenger deposit slashed to community pool.
+5. **Suspension and recovery**: Curator slashed below min bond → collection SUSPENDED → curator tops up bond → collection ACTIVE.
+6. **Suspension and closure**: Curator slashed → top_up_window expires → collection CLOSED → remaining bond refunded.
+7. **Featured batch**: Quality score ≥ 800, no challenges, active orders → featured for 7 days.
+8. **Price fairness**: Listing at median → 1000; listing 50% above median → 0.
+9. **Max collections**: Curator at max_collections_per_curator cannot create new collection.
+
+---
+
+## 13. Rollout plan
+
+| Phase | Scope | Dependencies |
+|-------|-------|-------------|
+| **v0 (advisory)** | AGENT-003 computes quality scores off-chain; KPI metrics in digest. No on-chain curation bonds or collections. | M010 (reputation), KOI MCP |
+| **v1 (on-chain curation)** | CosmWasm `marketplace-curation` contract with collection creation, bond lock, batch inclusion, challenge mechanism. Quality scores submitted on-chain. | M010, AGENT-003, DAO DAO |
+| **v2 (full marketplace)** | Curation fee routing via M013, featured batches, collection performance reports, basket token support. | M013, x/ecocredit/basket |
+
+---
+
+## Appendix: Source anchors
+
+| Document | Section |
+|----------|---------|
+| `phase-2/2.1-token-utility-mechanisms.md` | M011 Marketplace Curation & Quality Signals (lines 650–908) |
+| `phase-3/3.1-smart-contract-specs.md` | CosmWasm Contract: Marketplace Curation (lines 1080–1155) |
diff --git a/mechanisms/m011-marketplace-curation/datasets/README.md b/mechanisms/m011-marketplace-curation/datasets/README.md
@@ -0,0 +1,11 @@
+# m011 replay datasets
+
+Fixture files for replay testing of m011 (Marketplace Curation & Quality Signals) computations.
+
+## Files
+- `schema.json` — JSON Schema for the replay dataset format.
+- `fixtures/v0_sample.json` — Five scored credit batches across carbon (C), biodiversity (BT), and Kasigau (KSH) types, with three curated collections (ACTIVE and CLOSED). Quality scores match reference-impl output.
+- `fixtures/v0_collection_sample.json` — Three collections covering challenge lifecycle scenarios: UNDER_REVIEW (pending), SUSPENDED (challenger wins), ACTIVE (curator wins, challenge resolved).
+
+## Usage
+Feed fixture files into `m011_kpi.js` to verify KPI computation. Quality scores in fixtures correspond to `m011_score.js` output for the matching factor inputs.